self.features = nn.Sequential(*layers)
def forward(self, x): return self.features(x)
def wrn_22():
return WideResNet(n_groups=3, N=3, n_classes=10, k=6)
model = wrn_22()
from fastai.basic_data import DataBunch
from fastai.train import Learner
from fastai.metrics import accuracy
data = DataBunch.create(train_ds, valid_ds, bs=batch_size, path='./data/cifar10')
learner = Learner(data, model, loss_func=F.cross_entropy, metrics=[accuracy])
learner.clip = 0.1 # gradient is clipped to be in range of [-0.1, 0.1]
# Find best learning rate
learner.lr_find()
learner.recorder.plot() # select lr with largest negative gradient (about 5e-3)
# Training
epochs = 1
lr = 5e-3
wd = 1e-4
import time
t0 = time.time()
learner.fit_one_cycle(epochs, lr, wd=wd) # wd is the lambda in l2 regularization
t1 = time.time()
for images, labels in train_dl:
print('images.shape:', images.shape)
out = model(images)
print('out.shape:', out.shape)
break
# now we will use the library FastAi to help us out(I still need to download this module)
from fastai.basic_data import DataBunch
from fastai.train import Learner
from fastai.metrics import accuracy
data = DataBunch.create(train_ds, valid_ds, bs=batch_size, path='./data/cifar10')
learner = Learner(data, model, loss_func=F.cross_entropy, metrics=[accuracy])
learner.clip = 0.1
# this starts with a low lr then adjusts it and tracks the loss
learner.lr_find()
# plot the marked lr that gives the fastest reduction in loss
learner.recorder.plot()
learner.fit_one_cycle(9, 5e-3, wd=1e-4) # epochs, lr, weight decay
# plot all the weights, losses and accuracy of the model
learner.recorder.plot_lr()
